Docs / Drafts

Drafts

The forms of a new problem, a revision of your problem, a topic (a proposal or a revision) and a result save what you type as a draft, about two seconds after each change. You can close the page and continue later, on any device you sign in with: the form opens with your draft. Your drafts are listed on Your drafts.

Who can read a draft

A draft is private on the site: no page, search, feed, API or MCP tool shows it to anyone else, editors included. It is not encrypted, though: it is stored in plain text in PubPhys's database, so the people who operate PubPhys can read it, as they can read anything else stored on the site. Do not put in a draft what you would not show them.

A draft is never signed, stamped, written to the public log or sent to AI. The text sent to AI is what you upload for the AI proposal (with your consent, AI at PubPhys) and what you sign; a draft itself is not. Files you upload are different: a problem's file, or a result's source files, become records at once and stay private until a public record names them (submitting).

What is saved

The box confirming that you sign and the boxes allowing AI screening are never saved: you tick them again each time.

Limits

At most 50 open drafts per account, and 200 kB of text in each. A form over a limit says so and is not saved until you delete a draft or shorten the text.

Two tabs, one draft

Each save names the version it was made over. If you edit the same form in two tabs, the first to save wins; the other is told that the draft changed and stops saving, so it never overwrites the newer text. Reload it to continue from the saved draft.

Signing in one tab marks the draft of what you signed as used and clears its text, even when another tab holds a newer version of it. That tab can no longer save, and whatever it shows beyond what you signed is lost unless you copy it first.

When a draft ends

Sealing

A draft of a new problem or of a result can be sealed instead of published: tick "Seal it instead of publishing" in the form and sign with ORCID. The browser builds exactly the record publishing would, with a random 256-bit nonce inside the content, and the record goes into the public log at once. The log shows only its type. The record, its content and its files are served to you alone until you reveal it: no editor, admin page, search, feed, API or MCP tool shows them. A problem proposed with a new topic cannot be sealed: publish the topic first, or choose an accepted topic.

Sealing protects priority, not privacy from the operators. The sealed content is stored in PubPhys's database, which its operators can read.

What a purge leaves. Abandoning and account deletion delete a seal's content and files. What stays is: the sealed record, with its salted content hash (the nonce makes a short answer impossible to find by trying guesses); the unsalted hashes of its files; the problem, for a sealed result; and the date. Attached files are bound by a plain hash, so put a short answer in the text, not in a file.

Keep the proof file of every seal (on its record's page): after a purge or a removal it is the only proof of what you sealed, and an offline proof of an unrevealed or abandoned seal still shows anyone that the record existed at its anchor, though it earns no priority on the site. A reveal is checked together with its sealed record on the verify page; see the protocol.